The interconnectedness of gene regulation, protein interaction, and metabolic networks is responsible for the remarkable efficiency and adaptability of biological systems, as well as the extraordinary challenges facing researchers trying to understand them. The torrents of new biological data generated daily should lead to overcoming the challenges to understanding biological processes. However, our understanding of these systems has not grown proportionally to the amount of data generated. This disparity arises from the fact that the behavior of a biological system is not a linear superposition of the behaviors of its components. Higher-level structures within organisms can be maintained precisely because of the complex network of nonlinear interactions among lower-level components. As a result, scientists increasingly recognize that in order to advance our ability to understand and purposefully manipulate biomedical systems, we must take a systems level approach. However, it is not yet clear what systems level approach is optimal. I contend that we will only be able to make sense of systems-level information if we can develop methods that enable us to extract the small set of information that is relevant at the scale of interest.